AI Tech Talks Series
The SAFE AI Foundation hosts a series of technical talks on AI. Speakers are usually invited by us and the topics covered include AI models, security, governance, ethics, benchmarking, applications, etc.
Attendees must first register online and a confirmation will be sent to you after registration. All talks are free of charge and they last for usually 1 hour. Q&A is allowed at the end of the talk or whenever time permits. Come join us and it is also a good opportunity for you to network with the AI Community.
If you wish to speak at our event, please contact us via email with information (title, abstract, bio, date) about your talk. We will review and let you know if your proposal is acceptable. Email: contact@safeaifoundation.com
Date: 22 MAY 2026
( Registration is now OPEN )
Topic: AI Safety in Personal Banking: Risks and Remediations
Speaker: Mahalakshmi Venkateswarlu
Venue: Mission Branch, Santa Clara Library
Time: 12:00noon to 2:00pm (Talk + Q&A)
ABSTRACT
The rapid integration of Artificial Intelligence into banking has fundamentally shifted the landscape of financial security and consumer agency. While industry discourse often focuses on external threats like deepfakes, a more insidious vulnerability is emerging: the AI systems themselves. This talk pulls back the curtain on the modern financial ecosystem to explore the dual-edged nature of automated banking.
We begin by examining the "Inside-Out" threat, demonstrating how prompt injection and adversarial manipulation can turn a bank’s own defensive AI into an attack surface. Moving from security to equity, the session addresses the "Explainability Gap" in AI-driven lending—exploring how historical biases become hardcoded into decision-making engines, often leaving both regulators and consumers in the dark.
Drawing on primary research benchmarking LLM-based agents, this presentation provides a framework for evaluating AI reliability. Attendees will learn to distinguish between tasks suitable for automation and those requiring human oversight. The session concludes with a practical roadmap for the informed consumer, highlighting legal rights, red-flag signals, and the essential questions every customer should ask their financial institution before trusting an algorithm with their financial future.
REGISTRATION: Limited to first 40 attendees but you must first register at: https://safeaifoundation.com/form-register
BIO: Mahalakshmi is a Software and ML Engineer with six years of industry experience across Zoho Corporation and Cognizant Technology Solutions. Her current focus is on LLM systems, inference optimization, and AI agent reliability. She is currently pursuing an MS in Computer Science with an ML specialization at Georgia Tech's OMSCS program. Her most recent research "Less is More: Benchmarking LLM Based Recommendation Agents" was accepted for oral presentation at the LARS Workshop at the ACM Web Conference (WWW 2026). She has also served as a reviewer for PEARC26, and has been writing technical content on LLM inference, GPU systems for a practitioner audience.
